You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@ignite.apache.org by "Semen Boikov (JIRA)" <ji...@apache.org> on 2017/06/23 11:10:00 UTC

[jira] [Resolved] (IGNITE-5566) Improve TcpCommunicationSpi protocol for node stopping

     [ https://issues.apache.org/jira/browse/IGNITE-5566?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Semen Boikov resolved IGNITE-5566.
----------------------------------
    Resolution: Fixed
      Assignee:     (was: Semen Boikov)

Did these changes:
 - do not close connections in onContextDestroyed0 (it is called before discovery stop so remote nodes still consider node as alive and try to reconnect) 
- when node started stopping, then on incoming connections open reply with special message so that remote nodes do not retry reconnects


> Improve TcpCommunicationSpi protocol for node stopping
> ------------------------------------------------------
>
>                 Key: IGNITE-5566
>                 URL: https://issues.apache.org/jira/browse/IGNITE-5566
>             Project: Ignite
>          Issue Type: Improvement
>          Components: general
>            Reporter: Semen Boikov
>             Fix For: 2.1
>
>
> Currently if node starts stopping then TcpCommunicationSpi starts to close all incoming connections. In some benchmarks with large nodes number I observe that when client starts to stop it receives 1000s incoming connections from servers trying to connect to it. Need improve communication protocol and if node started to stop it should either immediately close NioServer or send some reply to connecting node so that it should stop connection attempts.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)